Hi Fausto! :)

see Architecture.txt No explicit locks are used. Due to the usage of a "ticketing" system, the progress of the requested operation must be polled.

Unfortunately, I use the terms "Task" and "Service" as a synonym. Task is meant to be an explicit instantiation of a service operation.

• Client sends a POST request to http://localhost:5000/request with attached form data like this: -- service : "InnerEarSegmentation" -- payload : data.tar.gz and receives -- task id -- access token to query the task's status/result.
• Server extracts the data and puts the request to the shortest queue
• Server performs operation and stores result
• Client queries http://host:5000/result/id&token and receives the result as a tar.gz file. The client can query the current status at http://host:5000/status/id&token The client can query the host capabilities including their interfaces at http://host:5000/supportedservices

The data is transfered via a single .tar.gz file during upload and download. The actual data is stored inside the .tar.gz file as -a file, probably with the following pattern: file-dataid.* -inside the main.json inside root/inputs/dataid The input data can be composed by defined input classes which perform a lot of simple error handling and extract the actual data. Currently implemented input classes are:

• VolumeInput
• FloatInput
• StringInput
• IntInput
• SingleChoiceInput These input classes also render the interfaces at http://host:5000/supportedservices.

The TaskScheduler takes care of assigning incoming tasks to different compute sources (GPUs). You can assign a set of possible tasks to each compute source individually.

Services can be defined easily. (e.g. see CustomTasks/FloatComputationTask.py) The given task_identifier has to match to the "service" parameter of the form of the initial POST request.

see all options in TOMAATSettings.py

Make sure the folder BASE_STORAGE_DIR defined in TOMAATSettings exists. Get a PostgreSQL database running with

• username: tomaat-user
• password: tomaat-password
• database: tomaat then...
• Run... python3 QueueManager.py python3 WebService.py python3 TaskScheduling.py

After Get It Running, using Chrome postman, you can perform a POST request to http://localhost:5000/request with body -> form-data and form entries:

• service: FloatComputation
• payload: payload_float_add.tar.gz

Get the result at http://host:5000/result/id&token

Yes, we should reorganize that.